Should baseline-dependent cut-offs really be used to define disease improvement in juvenile idiopathic arthritis? And few other considerations.

Rheumatology key message: Analysis of covariance should be favoured to responder analysis based on current baseline-dependent cut-offs for JIA improvement
Sir, In the recently published guideline of the European Medicines Agency on JIA [1], the different versions of the Juvenile Arthritis Disease Activity Score (JADAS) are quoted as valid primary end points to evaluate drug efficacy. Several sets of cut-offs are available to classify patients according to their disease activity status using the JADAS [2, 3]. In order to define disease improvement, cut-offs exist also for the change score (the difference between the JADAS value at baseline and follow-up) [3, 4].
An interesting example is the article by Horneff and Becker [4], in which cut-offs for the JADAS10 change score depending on groups of JADAS10 baseline values were developed. They assumed that ‘the seriously ill will regard only a considerable decrease in JADAS as an improvement, while less ill patients will perceive a less substantial decrease as an improvement’.
They concluded that the higher the baseline values, the greater the cut-off that satisfactorily separates patients whose disease will improve from those whose disease will not: for baseline JADAS10 groups [5,15), [15,25) and [25,40], the cut-offs were, respectively, 4, 10 and 17. Therefore, in a clinical trial, the responders’ rate would have …